Comparing QMT1 and HMMs for the synthesis of American English prosody

Three models are compared for the duration and pitch contour of American English in a speech synthesis framework. These models combine duration prediction by Quantification Metho d Type 1 (QMT1), a Codebook-based method for the F0 contour and a Hidden Markov Model-based method for both durations and F0. Subjective listening tests show that the HMMs are pre- ferred over the Codebook for the F0 contour, but that their dura- tion modelling performances are not significantly differen t from those of QMT1 in the tested setup. An analysis of naive free- form listener comments supports this fact, and suggests that such comments can give useful hints regarding the performance of each system.